Episodes
- SE0 — Series Overview — The New Data Unit for Coding Agents
- T1E0 — Topic 1 Overview — Foundations of Agentic Coding
- T1E1 — From Code Generation to Agentic Work
- T1E2 — The Agent-Computer Interface
- T1E3 — Trajectories as Training Data
- T1E4 — SWE-bench — Real Issues, Real Repositories
- T1E5 — SWE-agent and the Agent-Computer Interface
- T1E6 — Aider Benchmark — Can the Model Fix Its Own Mistakes?
- T1E7 — GitHub Copilot Cloud Agent — From Session to Reviewable Work
- T2E0 — Evaluation and Code Review for Coding Agents
- T2E1 — How to Evaluate a Coding Agent in Practice
- T2E2 — Public Tests, Hidden Tests, and Oracle Tests
- T2E3 — Leaderboard Literacy
- T2E4 — Code Review Agents: The Other Half of Agentic Coding
- T2E5 — Evaluating Review Quality
- T2E6 — Code Review Benchmarks and Datasets
- T2E7 — SWE-bench Verified
- T2E8 — SWE-bench-Live
- T2E9 — Aider Polyglot Benchmark
- T2E10 — Terminal-Bench 2.0
- T2E11 — FeatureBench
- T2E12 — CodeClash
- T2E13 — LLM-as-a-Verifier
- T2E15 — CodeReviewer
- T2E16 — CodeReviewQA
- T2E17 — CodeFuse-CR-Bench
- T2E18 — CR-Bench: Evaluating the Real-World Utility of AI Code Review Agents
- T2E19 — SWE-PRBench
- T2E20 — CodeReviewBench: Planted Bugs, Two Judges
- T2E21 — Martian's Code Review Bench: Graded by Real Developer Behavior
- T2E22 — MetaMateCR: AI-Assisted Fixes to Code Review Comments at Scale
- T2E23 — RepoQA: A Unit Test for Find-the-Code, Not Understand-the-Codebase
- T3E0 — Agent Systems, Harnesses, and Observability
- T3E1 — GitHub Copilot Coding Agent: From Task to Reviewable Work
- T3E2 — OpenAI Codex: Tasks, Workflows, and the Agent-First Loop
- T3E3 — Building a Multi-Agent Research System
- T3E4 — Effective Harnesses for Long-Running Agents
- T3E5 — Harness Design for Long-Running Application Development
- T3E6 — Harness Engineering in an Agent-First World
- T3E7 — Meta-Harness: End-to-End Optimization of Model Harnesses
- T3E8 — The Harness Problem
- T3E9 — Harness Engineering as a Discipline
- T4E0 — Training Signals and Improvement Studies
- T4E1 — CWM — Code Generation with World Models
- T4E2 — Debugging Code World Models
- T4E3 — SWE-Gym — Training Agents and Verifiers
- T4E4 — SWE-smith — Scaling Data for SWE Agents
- T4E5 — SWE-RL — Reinforcement Learning on Software Evolution
- T4E6 — Agent-RLVR — Guidance and Environment Rewards
- T4E7 — R2E-Gym — Procedural Environments and Hybrid Verifiers
- T4E8 — SWE-Hub — A Production System for Executable SWE Tasks
- T4E9 — Nemotron-Terminal — Data Engineering for Terminal Capability
- T4E10 — SWE-Bench Mobile
- T5E0 — Data Products for Agentic Coding Training
- T5E1 — The Agentic Coding Data Record
- T5E2 — Trajectory Data for Training and Analysis
- T5E3 — Verification and Review Labels
- T5E4 — From Production Telemetry to Training Data
- T5E5 — Governance and Splits
- T5E6 — The Six-Layer Content Schema